Spaces:
Running
Running
Create app.py
Browse files
app.py
ADDED
@@ -0,0 +1,89 @@
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
|
1 |
+
import gradio as gr
|
2 |
+
import requests.exceptions
|
3 |
+
from huggingface_hub import HfApi, hf_hub_download
|
4 |
+
from huggingface_hub.repocard import metadata_load
|
5 |
+
|
6 |
+
app = gr.Blocks()
|
7 |
+
|
8 |
+
def load_agent(model_id_1, model_id_2):
|
9 |
+
"""
|
10 |
+
This function load the agent's video and results
|
11 |
+
:return: video_path
|
12 |
+
"""
|
13 |
+
# Load the metrics
|
14 |
+
metadata_1 = get_metadata(model_id_1)
|
15 |
+
|
16 |
+
# Get the accuracy
|
17 |
+
results_1 = parse_metrics_accuracy(metadata_1)
|
18 |
+
|
19 |
+
# Load the video
|
20 |
+
video_path_1 = hf_hub_download(model_id_1, filename="replay.mp4")
|
21 |
+
|
22 |
+
# Load the metrics
|
23 |
+
metadata_2 = get_metadata(model_id_2)
|
24 |
+
|
25 |
+
# Get the accuracy
|
26 |
+
results_2 = parse_metrics_accuracy(metadata_2)
|
27 |
+
|
28 |
+
# Load the video
|
29 |
+
video_path_2 = hf_hub_download(model_id_2, filename="replay.mp4")
|
30 |
+
|
31 |
+
return model_id_1, video_path_1, results_1, model_id_2, video_path_2, results_2
|
32 |
+
|
33 |
+
def parse_metrics_accuracy(meta):
|
34 |
+
if "model-index" not in meta:
|
35 |
+
return None
|
36 |
+
result = meta["model-index"][0]["results"]
|
37 |
+
metrics = result[0]["metrics"]
|
38 |
+
accuracy = metrics[0]["value"]
|
39 |
+
return accuracy
|
40 |
+
|
41 |
+
def get_metadata(model_id):
|
42 |
+
"""
|
43 |
+
Get the metadata of the model repo
|
44 |
+
:param model_id:
|
45 |
+
:return: metadata
|
46 |
+
"""
|
47 |
+
try:
|
48 |
+
readme_path = hf_hub_download(model_id, filename="README.md")
|
49 |
+
metadata = metadata_load(readme_path)
|
50 |
+
print(metadata)
|
51 |
+
return metadata
|
52 |
+
except requests.exceptions.HTTPError:
|
53 |
+
return None
|
54 |
+
|
55 |
+
|
56 |
+
|
57 |
+
|
58 |
+
with app:
|
59 |
+
gr.Markdown(
|
60 |
+
"""
|
61 |
+
# Compare Deep Reinforcement Learning Agents 🤖
|
62 |
+
|
63 |
+
Type two models id you want to compare or check examples below.
|
64 |
+
""")
|
65 |
+
with gr.Row():
|
66 |
+
model1_input = gr.Textbox(label="Model 1")
|
67 |
+
model2_input = gr.Textbox(label="Model 2")
|
68 |
+
with gr.Row():
|
69 |
+
app_button = gr.Button("Compare models")
|
70 |
+
with gr.Row():
|
71 |
+
with gr.Column():
|
72 |
+
model1_name = gr.Markdown()
|
73 |
+
model1_video_output = gr.Video()
|
74 |
+
model1_score_output = gr.Textbox(label="Mean Reward +/- Std Reward")
|
75 |
+
with gr.Column():
|
76 |
+
model2_name = gr.Markdown()
|
77 |
+
model2_video_output = gr.Video()
|
78 |
+
model2_score_output = gr.Textbox(label="Mean Reward +/- Std Reward")
|
79 |
+
|
80 |
+
app_button.click(load_agent, inputs=[model1_input, model2_input], outputs=[model1_name, model1_video_output, model1_score_output, model2_name, model2_video_output, model2_score_output])
|
81 |
+
|
82 |
+
examples = gr.Examples(examples=[["sb3/a2c-AntBulletEnv-v0","sb3/ppo-AntBulletEnv-v0"],
|
83 |
+
["ThomasSimonini/a2c-AntBulletEnv-v0", "sb3/a2c-AntBulletEnv-v0"],
|
84 |
+
["sb3/dqn-SpaceInvadersNoFrameskip-v4", "sb3/a2c-SpaceInvadersNoFrameskip-v4"],
|
85 |
+
["ThomasSimonini/ppo-QbertNoFrameskip-v4","sb3/ppo-QbertNoFrameskip-v4"]],
|
86 |
+
inputs=[model1_input, model2_input])
|
87 |
+
|
88 |
+
|
89 |
+
app.launch()
|